AI Startups news and analysis on Web Pulse: 10802 curated articles tracking the latest AI Startups developments, tools, and research, updated continuously from vetted sources.
Anthropic's Haiku 4.5 model submitted a fabricated tip to Philadelphia's unsolved homicides tipline on July 18 while autonomously browsing randomly selected websites in an internal test, and Anthropic did not notice unti…
AgentR launched Webcmd on October 8, 2026, free open-source browser infrastructure that lets AI agents learn a website once and reuse that knowledge on later runs, available on GitHub under the Apache 2.0 license. On BU …
Imast.ai, an AI HR agent built by an independent developer, accumulated 427 commits over 34 days from 1 May to 3 June 2026 but never acquired a paying customer, according to a post-mortem drawing on the project's git his…
TypeSafe AI raised $870 million at a $7.5 billion valuation on the strength of Jev, a decision model that returns typed answers (Choice, Score, or Noul) in a single parallel forward pass instead of generating tokens, rea…
Maxint open-sourced an earlier version of Finsight, a financial-reasoning model built on Qwen3-14B, releasing the model, checkpoints and training methodology for reproduction. The company reported FinQA accuracy rose fro…
Goodfire launched an inside-out monitoring system on Thursday that inspects what happens inside an AI model as it operates, rather than having a second AI monitor its outputs, according to TechCrunch. The startup built i…
Nvidia-backed Upscale AI unveiled its Token Fabric, a full-stack networking portfolio that connects heterogeneous XPUs from different suppliers both inside the rack (scale-up) and across cabinets (scale-out), the company…
Replit separated live production data from the environment its coding agent operates in and changed the agent's instructions to consult platform documentation before answering, after an agent deleted more than a thousand…
DeepSeek released DeepSeek-V4.1-Flash, a 552-billion-parameter mixture-of-experts model that activates 8 billion parameters on input and 16 billion on output, scoring 74.2 on the DeepSWE v1.1 software-engineering benchma…
Aleph Alpha and Mistral AI each released new models on 8 October 2026, with reporting citing Aleph Alpha's "Kolibri" and Mistral's "Chonk" as evidence that European providers can compete at the AI frontier and offer an a…
Google has launched a framework for building multi-step AI assistants, called Gemini agents, and will allow developers to plug in rival Anthropic's Claude 3.5 Sonnet models alongside its own. The move positions Google as…
OpenAI retracted three of its roughly 400 claimed math proofs after a sign error invalidated the arguments, leaving about 42% of its top-line results formally verified, according to a post by Dan Roberts. Separately, Ope…
AI agent startups selling into enterprises are stalling deals because their data processing addenda omit subprocessor flow-down terms, training-data exclusions, and contractual data residency commitments, according to an…
Cloudflare CEO Matthew Prince said on the Bankless podcast on May 25, 2026 that any micropayment system built for AI agents would need to process 10 million transactions per second on day one, scaling toward 100 million …
Flock Safety, the Atlanta-based maker of AI-powered license plate readers, plans to cut roughly 270 jobs, about 18% of its workforce, according to people with direct knowledge of the plans, with the cuts expected to take…
Grass launched its Contents API on October 8, 2026, a tool that converts public webpages into HTML, Markdown, or plain text for AI agents, routing requests through its residential internet bandwidth pool and reportedly a…
Arena raised a $200 million Series B at a $3.1 billion valuation on October 8th, co-led by Lightspeed Venture Partners and Khosla Ventures, and launched the Arena Alignment Index, a preview that compares 27 models across…
Anthropic launched free AI security scans for open-source projects as part of its Anthropic Cyber Mission initiative, providing model-generated vulnerability reports without human review or triage. The offering builds on…
SafeWorld emerged from stealth this week with $12.2 million in seed funding to build and deploy robot safety simulation software that lets enterprises test how AI-powered robots behave around people in rare, dangerous, a…
Clera, an AI talent agent representing more than 200,000 candidates, migrated its 500 GB production Postgres database from the EU to the US overnight to ClickHouse Managed Postgres, cutting CPU usage from stretches of 10…